Google built on its launch of its Gemini large language model (LLM) with Gemini Pro, Google AI Studio and other tools developers and enterprises can use for generative AI use cases.
Last week, Google rolled out Gemini 1.0, its most powerful model designed to compete with OpenAI's ChatGPT. Gemini comes in three sizes with Gemini Ultra focused on complex tasks, Gemini Pro, an all-purpose model, and Gemini Nano, which is aimed at on-device usage.
With Google's follow up, the company looked to build out its AI stack and make it easier for developers and enterprises to build with Gemini. The plan for Google is to get Gemini into as many developer workflows as possible. With the introduction of Gemini, all three hyperscale cloud providers have announced or upgraded models in recent weeks. Amazon Web Services outlined Amazon Q at re:Invent. Microsoft is upgrading Copilot with the latest from OpenAI.
Among the key items:
- Gemini Pro is available to developers and enterprises with the Gemini Pro API available using Google AI Studio. Software developer kits for Gemini Pro are also available.
- Gemini Pro accepts text and generates text as output. Gemini Pro Vision is available also.
- Google AI Studio allows for 60 requests per minute. The company is looking to entice developers with more free offerings, an easy ability to transfer code and low costs once Gemini Pro is generally available in early 2024.
- Enterprises get the Gemini Pro API via Google AI Studio, which is available via Google Cloud Vertex AI.
- Vertex AI will enable developers to tune Gemini models with their own data and use a no-code, low-code environment.
- Gemini Pro is part of a curated list of more than 130 models available on Vertex AI.
- The Gemini Pro API is available on Vertex AI at no cost until general availability.
- Imagen 2, Google Cloud's text-to-image diffusion tool is available.
- MediLM, a healthcare industry LLM, is available via Google Cloud.
- Duet AI for Developers and Security Operations is now available. Duet AI for Developers includes AI code completion, code generation and chat in multiple integrated development environments. Tasks and processes are also streamlined. Duet AI for Security Operations aims to search data in seconds, add context and automate manual reviews.
- Mistral AI will use Google Cloud infrastructure to distribute and commercialize its LLMs. Mistral is a large AI provider in Europe.
With the moves, Google Cloud is building out its generative AI stack to better compete with the likes of Microsoft Azure and AWS. By offering AI Studio, a free web developer tool, Google Cloud is looking to entice more enterprises to build on its platform.
- AWS, Microsoft Azure, Google Cloud battle about to get chippy
- How much generative AI model choice is too much?
In a blog post, Google Cloud CEO Thomas Kurian said, "Gemini is part of a vertically integrated and vertically optimized AI technology stack."
This stack includes AI optimized infrastructure, models of different sizes, Vertex AI, and Duet AI assistive agents across Workspace and Google Cloud. Kurian said:
"The release of Gemini, Vertex AI and Duet AI offers a comprehensive and powerful cloud for developers and customers. Google Cloud is propelling the next generation of AI-powered agents across every industry."
Speaking during a Google Cloud webcast, Kurian said:
"Google Cloud Gemini Pro was designed to be our most efficient model to serve. So we priced it to be accessible to all developers. As for price with a focus on hardware and software optimization, we've reduced our pricing by four times per input character and two times per output character, at the same time, improving model quality and reducing serving latency."
Constellation Research analyst Holger Mueller said:
"Google Cloud lost no time to integrate Gemini across its offerings and did it in one week. Remember the time when it was one quarter for platform innovation to be uptaken? In the current AI race, Google does not want to lose any time with the first multi-modal GenAI LLM. The multimodal architecture of Gemini, coupled with the progress of the TPU V5 offerings, will likely prove to be the magic sauce for expanding Google's leadership in all things AI."
